METER-ML: A Multi-Sensor Earth Observation Benchmark for Automated Methane Source Mapping

Zhu, Bryan; Lui, Nicholas; Irvin, Jeremy; Lê, Jimmy T.; Tadwalkar, Sahil; Wang, Chenghao; Ouyang, Zutao; Liu, Frankie Y.; Ng, Andrew Y.; Jackson, Robert B.

doi:10.48550/arxiv.2207.11166

Cited by 2 publications

(3 citation statements)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Dataset. We select diverse RS datasets for providing a comprehensive assessment across different tasks within the RS field: seven RS classification datasets (namely AID [68], WHU-RS19 [13], NWPU [9], SIRI-WHU [86], EuroSAT [20], METER-ML [84], and fMoW [12]), two VQA datasets (LR and HR subsets of RSVQA [41]), and two visual grounding datasets (RSVG [56] and DIOR-RSVG [77]). We exclude any data from the test sets that overlap with the training set to avoid data leakage.…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

CMID: A Unified Self-Supervised Learning Framework for Remote Sensing Image Understanding

Muhtar

Zhang

Xiao

et al. 2023

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

Self-supervised learning (SSL) has gained widespread attention in the remote sensing (RS) and earth observation (EO) communities owing to its ability to learn task-agnostic representations without human-annotated labels. Nevertheless, most existing RS SSL methods are limited to learning either global semantic separable or local spatial perceptible representations. We argue that this learning strategy is suboptimal in the realm of RS, since the required representations for different RS downstream tasks are often varied and complex. In this study, we proposed a unified SSL framework that is better suited for RS images representation learning. The proposed SSL framework, Contrastive Mask Image Distillation (CMID), is capable of learning representations with both global semantic separability and local spatial perceptibility by combining contrastive learning (CL) with masked image modeling (MIM) in a self-distillation way. Furthermore, our CMID learning framework is architecture-agnostic, which is compatible with both convolutional neural networks (CNN) and vision transformers (ViT), allowing CMID to be easily adapted to a variety of deep learning (DL) applications for RS understanding. Comprehensive experiments have been carried out on four downstream tasks (i.e. scene classification, semantic segmentation, object-detection, and change detection) and the results show that models pre-trained using CMID achieve better performance than other state-of-the-art SSL methods on multiple downstream tasks. The code and pre-trained models will be made available at https://github.com/NJU-LHRS/official-CMID to facilitate SSL research and speed up the development of RS images DL applications.

show abstract

Section: Methodsmentioning

confidence: 99%

“…18. [41] VQA 500 UCM [47] Classification, Caption 2,519 RSVG [56] Visual Grounding 2,428 DIOR-RSVG [77] Visual Grounding 14,030 NWPU [9] Classification 4,941 METER-ML [84] Classification 1,400 RSITMD [76] Classification 504 fMoW [12] Classification 5,000 RSICD [42] Caption 1,000 Total 42,322…”

Section: H Evaluation Prompt For Each Taskmentioning

confidence: 99%

CMID: A Unified Self-Supervised Learning Framework for Remote Sensing Image Understanding

Muhtar

Zhang

Xiao

et al. 2023

IEEE Trans. Geosci. Remote Sensing

View full text Add to dashboard Cite

show abstract

“…However, to enable near-continuous monitoring of the world's oil and gas resources, it is essential that this process be supported by an automated detection and attribution system. Certain methods such as OGNET [2] and METER-ML [3] use deep neural networks to identify specific types of oil and gas site. This study focuses on the topic of automated detection of oil and gas infrastructures a topic that has not been explored in depth in the existing literature.…”

Section: Introductionmentioning

confidence: 99%

Oil and Gas Automatic Infrastructure Mapping: Leveraging High-Resolution Satellite Imagery Through Fine-Tuning of Object Detection Models

Guisiano,

Moulines,

Lauvaux

et al. 2023

Communications in Computer and Information Science

View full text Add to dashboard Cite

The oil and gas sector is the second largest anthropogenic emitter of methane, which is responsible for at least 25% of current global warming. To curb methane's contribution to climate change, emissions behavior from oil and gas infrastructure must be determined by an automated monitoring across the globe. This requires, as first step, an efficient solution to automatically detect and identify these infrastructures. In this extended study, we focus on automated identification of oil and gas infrastructure by using and comparing two types of advanced supervised object detection algorithms: Region-based Object Detector(YOLO and FASTER-RCNN) and Transformer-based Object Detector(DETR) with fine-tuning on our customized high-resolution satellite image database (Permian Basin U.S). The pre-training effect of each of these algorithms on detection results is studied and compared with non-pre-trained algorithms. The performed experiments demonstrate the general effectiveness of pre-trained YOLO v8 model with a Mean Average Precision over 90. The non-pre-trained model of this last one also over perform compare to FASTER-RCNN and DETR.

show abstract

METER-ML: A Multi-Sensor Earth Observation Benchmark for Automated Methane Source Mapping

Cited by 2 publications

References 0 publications

CMID: A Unified Self-Supervised Learning Framework for Remote Sensing Image Understanding

CMID: A Unified Self-Supervised Learning Framework for Remote Sensing Image Understanding

Oil and Gas Automatic Infrastructure Mapping: Leveraging High-Resolution Satellite Imagery Through Fine-Tuning of Object Detection Models

Contact Info

Product

Resources

About